Binary Information Press
In this paper, the authors present an approach using MapReduce paradigm to solve digraph problems, which can accommodate distributed data environment, such as cloud. The abstractions of both data and operations in origin version of MapReduce are too inexplicit to guideline the design and implement the corresponding algorithm. An extended variant of MapReduce with data structure and leveled operation abstraction is proposed to model the behaviors from users, programmers and system's perspectives. Two classic problems, layer digraph generation and max-flow, are chosen to exemplify how to use them, whereas the former is often taken as the foundational algorithm about other problems.